Transcript Religion nurtures some forms of prosocial behavior
The Nature and Nurture of Generosity: What can we learn from behavioral genetics?
René Bekkers
Center for Philanthropic Studies VU University Amsterdam
WIMPS, November 18, 2014
Thanks
• • • To the McArthur Foundation for funding the MIDUS data collection.
Colleagues who gave feedback: Dorret Boomsma, Dinand Webbink, Sara Konrath, Paul van Lange, Daniëlle Posthuma.
To be submitted as a chapter for a CESifo volume published at MIT Press.
November 18, 2014 WIMPS, IUPUI 2
November 18, 2014 WIMPS, IUPUI 3
November 18, 2014 WIMPS, IUPUI 4
November 18, 2014 WIMPS, IUPUI 5
Three questions
• • • How alike are twins in the United States with respect to prosocial behavior?
Are differences among twins in giving and volunteering related to differences in education and religion?
If so, what explains these relationships?
November 18, 2014 WIMPS, IUPUI 6
Prosocial behavior Formal: philanthropy Informal: helping
Money Time Social support
WIMPS, IUPUI November 18, 2014
Care
7
250 200 150 100 50 0 other natural sciences education and health public administration marketing & communication philanthropic studies economics sociology psychology
Number of publications per year about philanthropy by academic discipline (1899-2009)
Source: Bekkers & Dursun (2013), based on Bekkers & Wiepking (2011). ‘A Literature Review of Empirical Studies of Philanthropy: Eight Mechanisms that Drive Charitable Giving’. Nonprofit and Voluntary Sector Quarterly, 40(5): 924-973. Available at www.understandingphilanthropy.com
November 18, 2014 WIMPS, IUPUI 8
Ubiquitous correlates of philanthropy
1. Religion: – Affiliation (yes>no) – Denomination (Protestant>Catholic) – Participation (church attendance) 2. Education: – Level achieved
The variance between fields of study is small
November 18, 2014 WIMPS, IUPUI 9
Where do the correlations originate?
The more general research questions: 1. Why are religion and education correlated with prosocial behavior?
2. To what extent are these relationships the result of environmental influences?
3. Are these relationships causal?
November 18, 2014 WIMPS, IUPUI 10
Selection and causation
Behavior Education IQ, Other factors November 18, 2014 WIMPS, IUPUI 11
All selection, no causation
Behavior Education IQ, Other factors November 18, 2014 WIMPS, IUPUI 12
Selection and Causation
About 60%?
Education Factors X 1…n IQ, parental income, social science classes, college plans, extraversion, openness to experience Participation Bekkers, R. & Ruiter, S. (2008). ‘Education and voluntary association participation: Evidence for selection and causation’. Paper presented at the 103d ASA Annual Meeting, Boston, August 2, 2008.
November 18, 2014 WIMPS, IUPUI 13
Selection, causation, mediation
Mediating variable Education Behavior IQ, other factors November 18, 2014 Another mediating variable WIMPS, IUPUI 14
November 18, 2014 WIMPS, IUPUI 15
Three ‘theories’ on philanthropy
Philanthropy varies between social groups 1. because the resources of group members vary; 2. because the social values of groups vary; 3. because members of different groups have different self-identities.
November 18, 2014 WIMPS, IUPUI 16
November 18, 2014 WIMPS, IUPUI 17
The ideal experiment would randomize education
November 18, 2014 VWO = higher secondary education (≤ gymnasium) VMBO = lower vocational education WIMPS, IUPUI 18
Monozygotic twins
November 18, 2014 WIMPS, IUPUI 19
The unique environmental influence of education
Note that shared environmental influences are also excluded by design in this analysis November 18, 2014 WIMPS, IUPUI 20
A
What behavioral geneticists do: the ACE model
Additive genetic effects Typically 40-60% C Common (shared) environmental effects E Unique (non-shared) environmental effects (including error) Typically less than 10% (often zero) Typically 30-50% November 18, 2014 WIMPS, IUPUI 21
Note
• The first law of behavior genetics, as formulated by Eric Turkheimer (2000): “All human behavioral traits are heritable” • Eva Krapohl in a recent interview in The
Atlantic :
“Heritability describes what is; it does not predict what could be” November 18, 2014 WIMPS, IUPUI 22
A
ACE mediated effects model
Religiousness C Prosocial behavior E Total effect on prosocial behavior
A
10.2
C
27.6
Mediated by religiousness 7.5 (73.5%) 13.6
(49.3%) Koenig et al., 2007; n= 165 MZ and 100 DZ twin pairs November 18, 2014 WIMPS, IUPUI
E
62.3
2.9
(4.7%) 23
Biometric model fitting
• • • • • Fit statistics of various biometric models are compared to identify the best-fitting model.
Models depend on assumptions such as the Equal Environments Assumption.
The EEA is often disputed theoretically.
Empirical tests show it is often violated.
The resulting bias, however, seems to be minor (see Felson, Soc.Sc.Res., 2014).
November 18, 2014 WIMPS, IUPUI 24
What molecular geneticists do
• • • Genome Wide Association Studies (GWAS): identify ‘candidate genes’ that could explain variance in some outcome variable.
Typically, individual genes like OXTR and DRD4 explain tiny fractions of variance (<1%).
Typically, all single nucleotide polymorphisms (SNPs) combined explain less variance (16% of education) than estimated in biometric models (35%) – ‘missing heritability’ problem.
November 18, 2014 WIMPS, IUPUI 25
Where is the social science?
• • • • • In the variance explained by shared and unique environmental factors.
Let us rule out genetic effects by looking at monozygotic twins only. Any difference between MZ twins must have roots in the unique environment.
This choice avoids problems with the EEA.
Note that MZ twins also share 100% of shared environmental effects.
November 18, 2014 WIMPS, IUPUI 26
The problem
• “…families whose unobservable characteristics cause them to have a high likelihood of volunteering are also more likely to educate their children, so the relationship between schooling and volunteering is just a correlation caused by an excluded common cause.” John Gibson (2001) November 18, 2014 WIMPS, IUPUI 27
This is not my idea
• • • In 2001, New Zealand economist John Gibson published a study of volunteering among 85 identical twin pairs.
Though education in the pooled sample is associated with more volunteering, pairwise comparisons reveal the opposite. The twin with more years of education was found to volunteer fewer hours.
November 18, 2014 WIMPS, IUPUI 28
The implication
• • • Genetic effects cause a positive association between education and volunteering.
Unique environmental effects of education on volunteering are negative in this sample.
One interpretation of the negative effect is that it is the result of the opportunity cost of volunteering, potentially amplified by a decision making process within the household.
November 18, 2014 WIMPS, IUPUI 29
Related literature
• The twin fixed effects model has been used in economics to estimate the influence of schooling on income since the 1970s (Behrman & Taubman, 1976; Ashenfelter & Kreuger, 1994; Ashenfelter & Rouse, 1998; Isacsson, 1999; Miller, Mulvey & Martin, 1995; Bonjour et al., 2003).
November 18, 2014 WIMPS, IUPUI 30
E
Environment mediation model
Religiousness Prosocial behavior Education Note that this is a unique environment mediation model November 18, 2014 WIMPS, IUPUI 31
The MIDUS data
• • • Two wave longitudinal panel survey on Midlife in the United States (1995 and 2005) sponsored by the McArthur Foundation.
The RDD sample selection procedure included twin screening questions.
Only English-speaking respondents aged 25-74 living in the US who were physically and mentally able to complete the interview were allowed to participate.
November 18, 2014 WIMPS, IUPUI 32
Assessing zygosity
November 18, 2014 WIMPS, IUPUI 33
Are twins different at all?
MZ DZ Yes – here’s the discordance table:
Education
55% 64%
Religious affiliation
50% 53% Proportions of respondents from the same twin pair not reporting exactly the same level of education and religious affiliation November 18, 2014 WIMPS, IUPUI 34
ACE model results
Education Strength of religiosity Frequency of church attendance Amount donated ($) Hours volunteered Financial assistance to friends / family Hours helping friends / family
A
29.8
22.8
33.7
15.8
17.7
26.6
C
38.6
32.7
46.7
E
31.5
39.3
53.3
66.3
84.2
82.3
73.5
November 18, 2014 WIMPS, IUPUI 35
ACE model for volunteer hours
84,2 November 18, 2014 WIMPS, IUPUI A C E 36
ACE model for volunteer hours
84,2 November 18, 2014 WIMPS, IUPUI A C E 37
ACE model for volunteer hours
84,2 November 18, 2014 WIMPS, IUPUI A C E 38
ACE model for volunteer hours
15,8 A C E November 18, 2014 WIMPS, IUPUI 39
ACE model for volunteer hours
15,8 A C E November 18, 2014 WIMPS, IUPUI 40
ACE model for volunteer hours
15,8 A C E 41 November 18, 2014 WIMPS, IUPUI
Remember
• •
“Heritability describes what is; it does not predict what could be”.
These are the results of educational careers and systems for those in midlife in the US.
“All human behavioral traits are heritable”.
25% is not much compared to 75% for IQ.
November 18, 2014 WIMPS, IUPUI 42
The higher educated give more
These differences are massive: amounts donated in the top category are nine times the amount donated in the lowest category November 18, 2014 WIMPS, IUPUI 43
The higher educated volunteer more
November 18, 2014 Again, large differences WIMPS, IUPUI 44
Informal prosocial behaviors
November 18, 2014 WIMPS, IUPUI Perhaps Americans with less education know more people in need of support?
45
The religious give more
November 18, 2014 Religious giving is included in this figure. Excluding donations to religion, the differences are much smaller.
WIMPS, IUPUI 46
The religious volunteer more
This figure includes volunteering for religious organizations.
November 18, 2014 WIMPS, IUPUI 47
Informal prosocial behaviors
Perhaps Americans who attend church less often know more people in need of financial assistance and support?
48 November 18, 2014 WIMPS, IUPUI
Two basic regression models
1. Between effects model: ignores the twin pair structure, replicates bivariate analyses. Includes genetic + environmental effects.
2. Within MZ twin fixed effects model: does the higher educated / more religious twin of an MZ pair give and volunteer more than the less educated / religious co-twin? Includes environmental effects only.
November 18, 2014 WIMPS, IUPUI 49
Educational gradients among MZ twins
0,40 0,30 0,20 0,10 0,00 -0,10 -0,20 -0,30 Giving high school or less Non-religious giving some college more than some college Volunteering Non-religious volunteering November 18, 2014 WIMPS, IUPUI 50
Giving by co-twin
0,40 0,30 0,20 0,10 0,00 -0,10 -0,20 -0,30 1 2 3 1. high school or less 1 2 2. some college 3
Co-twin education x R education
1 2 3 3. more than some college November 18, 2014 WIMPS, IUPUI MZ DZ 51
Education and giving
400 350 300 250 200 150 100 50 0 *** Between siblings, DZ twins, and MZ twins November 18, 2014 *** FE DZ twins *** p <.001
FE MZ twins WIMPS, IUPUI 52
Two further models
• • Reduced form within MZ twin model: excludes religious denomination dummies, retaining education, church attendance and strength of religiosity.
Mediated reduced form within MZ twin model: adds social responsibility, prosocial self-identity, household income, and assets.
November 18, 2014 WIMPS, IUPUI 53
Education and giving among MZs
180 160 140 120 100 80 60 40 20 0 *** *** Between Within Reduced within Mediated reduced within November 18, 2014 WIMPS, IUPUI Total amount donated excluding religion *** p <.001
54
Education estimates on total giving
250 200 150 100 50 0 Between November 18, 2014 Within Reduced within Mediated reduced within WIMPS, IUPUI -1SE Estimate +1SE 55
Education and volunteering
0 -2 -4 -6 -8 -10 10 8 6 4 2 *** November 18, 2014 *** Between Within Reduced within Mediated reduced within WIMPS, IUPUI hours volunteered excluding religious *** p <.001
56
200
Resources and volunteering
* * ** ** 150 * 100 College Wages *$100 Assets *$1k 50 0 -50 Between Sibs Between Between Sibs Between Sibs Between Sibs *** p <.001
** p <.01
* p <.05
November 18, 2014 WIMPS, IUPUI 57
Resources and volunteering
200 150 100 50 *** *** *** *** *** 0 -50 Between MZ Between Between MZ ** Between MZ Between MZ *** p <.001
** p <.01
* p <.05
November 18, 2014 WIMPS, IUPUI College Wages *$100 Assets *$1k 58
Resources and volunteering
100 80 60 *** *** *** * * 40 20 0 -20 -40 Between MZ Between MZ Between MZ Between MZ Between MZ *** p <.001
** p <.01
* p <.05
College Wages *$100 Assets *$1k November 18, 2014 WIMPS, IUPUI 59
Resources and volunteering
100 80 60 40 20 0 -20 -40 Fixed MZ Fixed MZ Fixed MZ Fixed MZ Fixed MZ *** p <.001
** p <.01
* p <.05
College Wages *$100 Assets *$1k November 18, 2014 WIMPS, IUPUI 60
Church attendance and giving
25 20 15 10 5 0 *** Between *** Within *** *** Amount excluding religion Reduced within Mediated reduced within WIMPS, IUPUI *** p <.001
November 18, 2014 61
Church attendance and volunteering
0 -0,2 -0,4 -0,6 -0,8 -1 0,8 1 0,6 0,4 0,2 Between November 18, 2014 Within Reduced within Mediated reduced within WIMPS, IUPUI hours volunteered excluding religious 62
Strength of religiosity and giving
300 250 200 150 100 50 0 -50 -100 Between November 18, 2014 * * WIMPS, IUPUI * Amount excluding religion Within Reduced within Mediated reduced within * p <.05
63
Religiosity and volunteering
15 10 5 0 40 35 30 25 20 ** ** Between *** *** Within *** *** *** Reduced within Mediated reduced within WIMPS, IUPUI hours volunteered excluding religious *** p <.001; ** p < .01
November 18, 2014 64
November 18, 2014 WIMPS, IUPUI 65
November 18, 2014 WIMPS, IUPUI 66
Conclusions
• • • The association between the level of education and giving and volunteering is due to genetic or shared environmental effects.
The association between religiosity and charitable giving is due to unique environmental effects, but it is limited to church contributions.
Religiosity nurtures volunteering, also beyond religious organizations.
November 18, 2014 WIMPS, IUPUI 67
Mediation
• • • • Education hardly mediates unique environmental influences on giving (-0.5%) or volunteering (1.8%).
Religion mediates unique environmental influences on giving (15.6%) but not on volunteering (2.0%).
Education effects are partly mediated (25%) by income and assets.
Religiosity effects are mediated by prosocial self-identity (55%), but not by prosocial values.
November 18, 2014 WIMPS, IUPUI 68
Or vice versa
• • • • Perhaps volunteering nurtures religiosity.
Or perhaps an omitted (shared?) environmental effect nurtures volunteering and religiosity.
We cannot infer causality from the twin fixed effects model.
But we can look at changes in religiosity and volunteering between the two waves.
November 18, 2014 WIMPS, IUPUI 69
…and?
• • Respondents who quit volunteering between the first and the second wave are less frequently attending church and report lower strength of religiosity in the second wave than in the first wave.
Respondents who started volunteering are more frequently attending church in the second wave. November 18, 2014 WIMPS, IUPUI 70
The measurement error problem
• • Could differential measurement error explain the pattern of results?
That is unlikely. The test-retest correlation of education is higher (.87) than that of the frequency of church attendance (.72). It is similar to strength of religiosity (.84).
November 18, 2014 WIMPS, IUPUI 71
The variance problem
• • Could a differential lack of variance explain the pattern of results?
That is unlikely. MZ twins are more likely to be discordant with respect to education (55%) than with respect to religion (50%).
72 November 18, 2014 WIMPS, IUPUI
Future directions
• • Replicate this finding using data from other samples of twins, in the US and beyond.
Examine other dependent variables using this method: trust, subjective well being, prosocial values...
73 November 18, 2014 WIMPS, IUPUI
Schooling effects on income
5 0 -5 35 30 25 20 15 10 *** November 18, 2014 *** Between *** *** Siblings FE *** *** DZ FE WIMPS, IUPUI Education (12 cat) College+ MZ FE *** p <.001
74
René Bekkers
Center for Philanthropic Studies VU University Amsterdam [email protected]
Blog: http://renebekkers.wordpress.com
Twitter: @renebekkers
References
• • • • • Bekkers, R. & Dursun, E. (2013). “A Brief History of Research on Philanthropy.” http://www.understandingphilanthropy.com
Felson, J. (2014). “What can we learn from twin studies? A comprehensive evaluation of the equal environments assumption.” Social Science Research, 43: 184-199.
Gibson, J. (2001). “Unobservable Family Effects and the Apparent External Benefits of Education.” Economics of Education Review, 20: 225-233.
Koenig, L.B., McGue, M., Krueger, R.F., Bouchard, T.J. (2007). “Religiousness, Antisocial Behavior, and Altruism: Genetic and Environmental Mediation.” Journal of Personality, 75: 265-290. Reuter, M., Felten, A., Penz, S., Mainzer, A., Markett, S. & Montag, C. (2013). “The influence of dopaminergic gene variants on decision making in the ultimatum game.” Frontiers in Human Neuroscience, 7: 1-8.
November 18, 2014 WIMPS, IUPUI 76
More references
• • • • • • • Ashenfelter, O., & Krueger, A. (1994). “Estimates of the economic return to schooling from a new sample of twins.” American Economic Review, 84, 1157–1173.
Ashenfelter, O., & Rouse, C. (1998). “Income, schooling and ability: Evidence from a new sample of identical twins.” Quarterly Journal of Economics, 113, 153–284. Behrman, J. & Taubman, P. (1976). “Intergenerational Transmission of Income and Wealth.” American Economic Review, 66: 436-440.
Behrman, J. & Rosenzweig, M.R. (1999). “Ability biases in schooling returns and twins: a test and new estimates.” Economics of Education Review, 18: 159-167.
Bonjour, D., Cherkas, L., Haskel, J., Hawkes, D., & Spector, T. (2003). “Returns to Education: Evidence from UK Twins.” American Economic Review, 93: 1799-1812.
Isacsson, G. (1999). “Estimates of the Return to Schooling in Sweden from a Large Sample of Twins.” Labour Economics, 6: 471-489.
Miller, P., Mulvey, C. & Martin, N. (1995). “What Do Twins Studies Reveal About the Economic Returns to Education? A Comparison of Australian and U.S. Findings." American Economic Review, 85: 586-599.
November 18, 2014 WIMPS, IUPUI 77
Measures
Donations. Donations to organizations were measured with the following question: “On average, about how many dollars per month do you or your family living with you contribute to each of the following people or organizations? If you contribute food, clothing, or other goods, include their dollar value. (If none, enter "0".)” After this introduction, donations to three categories of organizations were measured: (1) to religious groups; (2) to political organizations or causes; (3) to any other organizations, causes, or charities (including donations made through monthly payroll deductions)? Amounts donated per month were multiplied by 12 to obtain the total amount donated per year. The sum of these contributions is the variable for the total amount donated to organizations. A separate variable was created excluding donations to religion to see if the relationship between religion and philanthropy would also hold for ‘secular giving’. The test-retest correlation of the total amount donated measured in dollars is .25; for the logtransformed amounts the test-retest correlation is .44. For donations to organizations other than religion the test-retest correlation of the dollar amounts is .29; for the log-transformed amounts it is .39.
November 18, 2014 WIMPS, IUPUI 78
Volunteering. The questions on volunteering in M1 and M2 asked about four types of formal volunteer work: ‘hospital, nursing home, or other health care-oriented work’, ‘school or other youth-related volunteer work’, ‘volunteer work for political organizations or causes’, and ‘volunteer work for any other organization, cause or charity’. While these questions did not explicitly identify religious organizations, respondents could report volunteering for religious organizations in the question about any ‘other’ organizations. A separate variable was created excluding potentially religious volunteering by computing the sum of hours volunteered in the first three types. The test-retest correlation of the total number of volunteer hours is .38; for the log-transformed hours the test-retest correlation is .46. For the hours volunteered in organizations other than religious organizations the test-retest correlation is .28; for the log-transformed variable it is .36.
November 18, 2014 WIMPS, IUPUI 79
November 18, 2014 WIMPS, IUPUI 80
November 18, 2014 WIMPS, IUPUI 81
November 18, 2014 WIMPS, IUPUI 82
November 18, 2014 WIMPS, IUPUI 83
November 18, 2014 WIMPS, IUPUI 84
November 18, 2014 WIMPS, IUPUI 85